debug: (inference) sample script atleast ?!

inference.py

@@ -1,270 +1,188 @@
+"""
+Inference Script Example
+===================================
+MANDATORY
+- Before submitting, ensure the following variables are defined in your environment configuration:
+    API_BASE_URL   The API endpoint for the LLM.
+    MODEL_NAME     The model identifier to use for inference.
+    HF_TOKEN       Your Hugging Face / API key.
+    LOCAL_IMAGE_NAME The name of the local image to use for the environment if you are using from_docker_image()
+                     method
+
+- Defaults are set only for API_BASE_URL and MODEL_NAME 
+    (and should reflect your active inference setup):
+    API_BASE_URL = os.getenv("API_BASE_URL", "<your-active-endpoint>")
+    MODEL_NAME = os.getenv("MODEL_NAME", "<your-active-model>")
+    
+- The inference script must be named `inference.py` and placed in the root directory of the project
+- Participants must use OpenAI Client for all LLM calls using above variables
+
+STDOUT FORMAT
+- The script must emit exactly three line types to stdout, in this order:
+
+    [START] task=<task_name> env=<benchmark> model=<model_name>
+    [STEP]  step=<n> action=<action_str> reward=<0.00> done=<true|false> error=<msg|null>
+    [END]   success=<true|false> steps=<n> score=<score> rewards=<r1,r2,...,rn>
+
+  Rules:
+    - One [START] line at episode begin.
+    - One [STEP] line per step, immediately after env.step() returns.
+    - One [END] line after env.close(), always emitted (even on exception).
+    - reward and rewards are formatted to 2 decimal places.
+    - done and success are lowercase booleans: true or false.
+    - error is the raw last_action_error string, or null if none.
+    - All fields on a single line with no newlines within a line.
+    - Each tasks should return score in [0, 1]
+
+  Example:
+    [START] task=click-test env=miniwob model=Qwen3-VL-30B
+    [STEP] step=1 action=click('123') reward=0.00 done=false error=null
+    [STEP] step=2 action=fill('456','text') reward=0.00 done=false error=null
+    [STEP] step=3 action=click('789') reward=1.00 done=true error=null
+    [END] success=true steps=3 score=1.00 rewards=0.00,0.00,1.00
+"""
+
 import asyncio
-import json
 import os
-import sys
-import re
-from pathlib import Path
+import textwrap
 from typing import List, Optional
-import httpx
-from dotenv import load_dotenv
-from openai import OpenAI
-
-ROOT = Path(__file__).resolve().parent
-REPO_ROOT = ROOT.parent
-
-if str(REPO_ROOT) not in sys.path:
-    sys.path.insert(0, str(REPO_ROOT))
-
-load_dotenv()
 
-TASK_DIR = ROOT / "tasks"
-SYSTEM_PROMPT = (ROOT / "system_prompt.txt").read_text().strip()
-
-HF_TOKEN = os.getenv("HF_TOKEN") or os.getenv("API_KEY")
-MODEL_NAME = os.getenv("MODEL_NAME", "gpt-4o")
-
-BASE_URL = "https://vishvam10-cloudenv.hf.space"
-
-TASK_ORDER = ["easy-task", "medium-task", "hard-task"]
-
-MAX_PARSE_RETRIES = 3
-MAX_STEPS = 5
+from openai import OpenAI
 
-client = OpenAI(base_url="https://api.openai.com/v1", api_key=HF_TOKEN)
+from my_env_v4 import MyEnvV4Action, MyEnvV4Env
+IMAGE_NAME = os.getenv("IMAGE_NAME") # If you are using docker image 
+API_KEY = os.getenv("HF_TOKEN") or os.getenv("API_KEY")
+
+API_BASE_URL = os.getenv("API_BASE_URL") or "https://router.huggingface.co/v1"
+MODEL_NAME = os.getenv("MODEL_NAME") or "Qwen/Qwen2.5-72B-Instruct"
+TASK_NAME = os.getenv("MY_ENV_V4_TASK", "echo")
+BENCHMARK = os.getenv("MY_ENV_V4_BENCHMARK", "my_env_v4")
+MAX_STEPS = 8
+TEMPERATURE = 0.7
+MAX_TOKENS = 150
+SUCCESS_SCORE_THRESHOLD = 0.1  # normalized score in [0, 1]
+
+# Max possible reward: each token contributes 0.1, across all steps
+_MAX_REWARD_PER_STEP = MAX_TOKENS * 0.1
+MAX_TOTAL_REWARD = MAX_STEPS * _MAX_REWARD_PER_STEP
+
+SYSTEM_PROMPT = textwrap.dedent(
+    """
+    You are interacting with a simple echo environment.
+    Each turn you must send a message. The environment will echo it back.
+    Reward is proportional to message length: reward = len(message) * 0.1
+    Your goal is to maximize total reward by sending meaningful, substantive messages.
+    Reply with exactly one message string — no quotes, no prefixes, just the message text.
+    """
+).strip()
 
 
 def log_start(task: str, env: str, model: str) -> None:
     print(f"[START] task={task} env={env} model={model}", flush=True)
 
 
-def log_step(
-    step: int, action: str, reward: float, done: bool, error: Optional[str]
-) -> None:
+def log_step(step: int, action: str, reward: float, done: bool, error: Optional[str]) -> None:
     error_val = error if error else "null"
     done_val = str(done).lower()
-    action_safe = action.replace("\n", " ").replace("\r", "")[:120]
     print(
-        f"[STEP] step={step} action={action_safe} reward={reward:.2f} "
-        f"done={done_val} error={error_val}",
+        f"[STEP] step={step} action={action} reward={reward:.2f} done={done_val} error={error_val}",
         flush=True,
     )
 
 
-def log_end(
-    success: bool, steps: int, score: float, rewards: List[float]
-) -> None:
+def log_end(success: bool, steps: int, score: float, rewards: List[float]) -> None:
     rewards_str = ",".join(f"{r:.2f}" for r in rewards)
-    print(
-        f"[END] success={str(success).lower()} steps={steps} "
-        f"score={score:.3f} rewards={rewards_str}",
-        flush=True,
-    )
-
+    print(f"[END] success={str(success).lower()} steps={steps} score={score:.3f} rewards={rewards_str}", flush=True)
 
-def load_task(name: str):
-    return json.loads((TASK_DIR / f"{name}.json").read_text())
 
+def build_user_prompt(step: int, last_echoed: str, last_reward: float, history: List[str]) -> str:
+    history_block = "\n".join(history[-4:]) if history else "None"
+    return textwrap.dedent(
+        f"""
+        Step: {step}
+        Last echoed message: {last_echoed!r}
+        Last reward: {last_reward:.2f}
+        Previous steps:
+        {history_block}
+        Send your next message.
+        """
+    ).strip()
 
-async def register_pipeline(task: dict):
-    async with httpx.AsyncClient(timeout=30) as http:
-        r = await http.post(
-            f"{BASE_URL}/api/pipelines",
-            json={
-                "pipeline_id": task["pipeline_id"],
-                "broken_pipeline": task["broken_pipeline"],
-                "ideal_pipeline": task["ideal_pipeline"],
-            },
-        )
-
-        if r.status_code not in (200, 201, 400):
-            print(
-                f"[PIPELINE REGISTER ERROR] {r.status_code} {r.text}",
-                flush=True,
-            )
-
-
-async def env_reset(task_id: str, episode_id: str):
-    async with httpx.AsyncClient(timeout=30) as http:
-        r = await http.post(
-            f"{BASE_URL}/reset",
-            json={"task_id": task_id, "episode_id": episode_id},
-        )
-        r.raise_for_status()
-        return r.json()
 
-
-async def env_step(action: dict):
-    async with httpx.AsyncClient(timeout=30) as http:
-        r = await http.post(
-            f"{BASE_URL}/step",
-            json={"action": action},
+def get_model_message(client: OpenAI, step: int, last_echoed: str, last_reward: float, history: List[str]) -> str:
+    user_prompt = build_user_prompt(step, last_echoed, last_reward, history)
+    try:
+        completion = client.chat.completions.create(
+            model=MODEL_NAME,
+            messages=[
+                {"role": "system", "content": SYSTEM_PROMPT},
+                {"role": "user", "content": user_prompt},
+            ],
+            temperature=TEMPERATURE,
+            max_tokens=MAX_TOKENS,
+            stream=False,
         )
-        r.raise_for_status()
-        return r.json()
+        text = (completion.choices[0].message.content or "").strip()
+        return text if text else "hello"
+    except Exception as exc:
+        print(f"[DEBUG] Model request failed: {exc}", flush=True)
+        return "hello"
 
 
-def clean_json(raw: str) -> str:
-    raw = raw.strip()
-    raw = re.sub(r"^```(?:json)?\s*", "", raw)
-    raw = re.sub(r"\s*```$", "", raw)
+async def main() -> None:
+    client = OpenAI(base_url=API_BASE_URL, api_key=API_KEY)
 
-    match = re.search(r"\{.*\}", raw, re.DOTALL)
-    if match:
-        return match.group(0)
+    env = await MyEnvV4Env.from_docker_image(IMAGE_NAME)
 
-    return raw
+    history: List[str] = []
+    rewards: List[float] = []
+    steps_taken = 0
+    score = 0.0
+    success = False
 
+    log_start(task=TASK_NAME, env=BENCHMARK, model=MODEL_NAME)
 
-def build_action(raw: str, task_id: str, episode_id: str):
     try:
-        raw = clean_json(raw)
-        data = json.loads(raw)
-
-        if isinstance(data, dict) and "action" in data:
-            data = data["action"]
-
-        if not isinstance(data, dict):
-            return None
-
-        service = data.get("service")
-        operation = data.get("operation")
-
-        if not service or not operation:
-            return None
+        result = await env.reset() # OpenENV.reset()
+        last_echoed = result.observation.echoed_message
+        last_reward = 0.0
 
-        instance_id = data.get("instance_id")
+        for step in range(1, MAX_STEPS + 1):
+            if result.done:
+                break
 
-        if isinstance(service, str) and ":" in service:
-            service, parsed_instance = service.split(":", 1)
-            instance_id = instance_id or parsed_instance
+            message = get_model_message(client, step, last_echoed, last_reward, history)
 
-        return {
-            "task_id": task_id,
-            "episode_id": episode_id,
-            "service": service.lower(),
-            "instance_id": instance_id,
-            "operation": operation.strip(),
-            "payload": data.get("payload", {})
-            if isinstance(data.get("payload"), dict)
-            else {},
-            "principal_arn": data.get("principal_arn"),
-            "intent": data.get("intent"),
-        }
+            result = await env.step(MyEnvV4Action(message=message))
+            obs = result.observation
 
-    except Exception:
-        return None
+            reward = result.reward or 0.0
+            done = result.done
+            error = None
 
+            rewards.append(reward)
+            steps_taken = step
+            last_echoed = obs.echoed_message
+            last_reward = reward
 
-def call_model(step, state, broken, ideal, history):
-    prompt = f"""
-Return ONLY valid JSON.
-
-STEP {step}
-STATE {json.dumps(state)[:1000]}
-BROKEN {json.dumps(broken)[:1000]}
-IDEAL {json.dumps(ideal)[:1000]}
-HISTORY {history[-5:]}
-"""
-
-    res = client.chat.completions.create(
-        model=MODEL_NAME,
-        messages=[
-            {"role": "system", "content": SYSTEM_PROMPT},
-            {"role": "user", "content": prompt},
-        ],
-        temperature=0.3,
-        max_tokens=500,
-    )
-
-    return (res.choices[0].message.content or "").strip()
-
-
-async def run_episode(task):
-    task_id = task["pipeline_id"]
-    episode_id = f"{task_id}-episode-1"
-
-    log_start(task=task_id, env="http", model=MODEL_NAME)
-
-    result = await env_reset(task_id, episode_id)
-
-    await env_step(
-        {
-            "task_id": task_id,
-            "episode_id": episode_id,
-            "service": "internal",
-            "instance_id": None,
-            "operation": "InitializeEnvironment",
-            "payload": {},
-        }
-    )
-
-    await register_pipeline(task)
+            log_step(step=step, action=message, reward=reward, done=done, error=error)
 
-    obs = {}
-    history = []
-    rewards = []
-    step_count = 0
+            history.append(f"Step {step}: {message!r} -> reward {reward:+.2f}")
 
-    for step in range(MAX_STEPS):
-        if obs.get("done"):
-            break
+            if done:
+                break
 
-        error = None
+        score = sum(rewards) / MAX_TOTAL_REWARD if MAX_TOTAL_REWARD > 0 else 0.0
+        score = min(max(score, 0.0), 1.0)  # clamp to [0, 1]
+        success = score >= SUCCESS_SCORE_THRESHOLD
 
+    finally:
         try:
-            raw = call_model(
-                step,
-                obs,
-                task["broken_pipeline"],
-                task["ideal_pipeline"],
-                history,
-            )
-
-            action = build_action(raw, task_id, episode_id)
-
-            if action is None:
-                error = "invalid_action"
-                log_step(step, "INVALID_ACTION", 0.0, False, error)
-                continue
-
-            result = await env_step(action)
-            obs = result["observation"]
-
-            reward = float(obs.get("reward", 0.0))
-            done = bool(obs.get("done", False))
-
-            rewards.append(reward)
-            history.append(raw)
-            step_count += 1
-
-            log_step(
-                step=step_count,
-                action=str(action),
-                reward=reward,
-                done=done,
-                error=None,
-            )
-
+            await env.close()
         except Exception as e:
-            error = str(e)
-
-            log_step(step, "ERROR", 0.0, False, error)
-            break
-
-        if obs.get("done"):
-            break
-
-    success = obs.get("done", False)
-    score = sum(rewards) / (len(rewards) or 1)
-
-    log_end(success=success, steps=step_count, score=score, rewards=rewards)
-
-
-async def main():
-    for name in TASK_ORDER:
-        task = load_task(name)
-        await run_episode(task)
+            print(f"[DEBUG] env.close() error (container cleanup): {e}", flush=True)
+        log_end(success=success, steps=steps_taken, score=score, rewards=rewards)
 
 
 if __name__ == "__main__":
-    asyncio.run(main())
+    asyncio.run(main())